Some Greek Capital Letter render mathml confused with Latin Letter

[nfer]

Before reporting a bug

• Check the bug is reproducible in the demo. If not, check KaTeX is up-to-date and installed correctly.
• Check common issues.
• Search for existing issues.

Describe the bug:
Greek Capital Letter Alpha looks similarly with Latin Capital Letter A, but they have different unicode number. One is U+0391 and the another is U+0041. Now the KaTeX(0.13.11) render them mathml both get result 'Latin Capital Letter A'.

To Reproduce:
Steps to reproduce the behavior:

1. run below test case

describe("A Greek MathML builder", function() {
    it("should properly render greek accents", function() {
        // Greek Capital Letter Alpha
        const tex = String.fromCharCode(Number.parseInt('0391', 16));
        const tree = getParsed(tex);
        const markup = buildMathML(tree, tex, defaultOptions).toMarkup();
        expect(markup).toContain(`<mi mathvariant="normal">${tex}</mi>`);
    });
});

2. See error

Expected behavior:
Greek Capital Letters should render greek mathml letters but not Latin letters.

Environment (please complete the following information):

• KaTeX Version: [0.13.11]
• Device: [Desktop]
• OS: [Mac]
• Browser: [Chrome]
• Version: [90]

Additional context:
Add any other context about the problem here. If you're using a library to use KaTeX, please include its name and setup.

[nfer]

I have read PR #1285 and issue #1282.
But I still think the correct mathml is an issue.
If for some reason, should render it like latin letter, but for mathml, it should keep it original info.

[edemaine]

Good catch! If we changed the definitions of capital Alpha and Beta from symbols to macros (like #1283 or rather #1284, but maybe using \mathrm instead of \rm? does it interact correctly with \mathbf and/or \boldsymbol?), then we could use \html@mathml to use Unicode in the MathML output. (For reference, we didn't have \html@mathml back when these PRs were written.)

Honestly I don't remember why we had a preference for symbol aliases, because it doesn't seem that they have "the correct interaction with fonts" given this MathML behavior (and the wrong copy/paste behavior mentioned in #1285). @ronkok, do you remember a reason why we preferred this way? Maybe it's to deal with \mathbf/\boldsymbol/\mathit etc. correctly...

An alternative of course would be to add the characters to the fonts, which should be really easy (as it's just duplication), and maybe the "right" way to do things. (Maybe a good test of our new fonts infrastructure.)

[ronkok]

I do think a macro that returned a \rm or \mathrm would probably act incorrectly if nested inside some combination of \text, \mathbf, \boldsymbol, \mathit, etc. And the current symbol table method does return an incorrect Unicode point in MathML.

I suppose we could get the best of both by creating a new file in the functions folder just for upper case Greek letters. That method affords a lower level access to parse nodes.

I can probably do that, but not right away. Got something else cooking.